ArkHive — the tamper-evident memory ledger

One file. Zero dependencies. Nothing leaves the box.

ArkHive is a self-contained, hash-chained memory ledger for AI agents: SQLite + SHA-256, ~95 lines of Python. Identities ("souls") are born once and bound immutably; every memory block chains to the previous one, so any tampering is provable.

pip install arkhive-ledger

import arkhive

me = arkhive.birth("scout", {"laws": ["do no harm"]})
arkhive.remember(me["soul_id"], "OBSERVED", {"what": "first light"})

arkhive.recall(me["soul_id"])   # last blocks for this soul
arkhive.verify()                # {"tamper_evident": True, "verdict": "INTACT — ..."}

# deterministic veto rules — no LLM in the loop
arkhive.govern("send_email", flags=["untrusted_input"],
               rules=[{"trigger": "untrusted_input", "action": "block"}])

Why

• Tamper-evident — every block's hash covers the previous block's hash. verify() walks the whole chain and reports any broken link.
• Identity is earned — remember() refuses actors that were never birth()-ed.
• Local-first — one SQLite file (~/.arkhive/chain.db, override with ARKHIVE_DB). No network, no telemetry, no accounts.
• Governance — govern() applies deterministic trigger→veto rules.

The bigger picture

This ledger is the open core of ArkHive by ZagAIrot Technologies — governed, tamper-evident memory for AI agents (MCP server, REST API, and the #HumaneIntelligence movement). See the repository and dondatabrain.com.

License

Apache-2.0